쉐도잉 연습: Is AI the next dot-com crash? | Business Beyond - YouTube로 영어 말하기 배우기

Billions are being poured into data centers to make two magical letters work.

⏸ 일시 정지

재생 속도:

298 문장

문장이 너무 짧거나 길면 Edit를 눌러 조정하세요.

Billions are being poured into data centers to make two magical letters work.

AI.

Artificial intelligence.

The one technology to save us all.

You shouldn't be asking what it can do,

but what it can do if you buy into the hype.

This technology has extremely high narratability.

We might have AI that is smarter than any human by the end of this year.

I have engineers within Anthropoc who say I don't write any code anymore.

But at the same time,

headlines like 95% of companies don't report profit increases connected to AI make you wonder,

is the hype actually legit or are we really close to an inflated bubble?

To get to the limits and problems of AI

and its bubbly character

that I'm going to explore with these two we first need to get what companies can actually use AI for

and that's why I went to BMW.

It's not that BMW builds entire cars with AI yet

but they're using AI to manage their production equipment which they hand over to production partners.

That can be presses, welding machines,

lifting cranes, you name it and thousands of them.

AI helps to keep track of these 250 000 tools that BMW needs to manage.

Our AI is like small little helpers which do specific tasks.

So they are creating the inventory order,

they are sending the tasks to the supplier,

they are checking the results and at the end of the day if everything is green,

so green light, you say okay this task can go to our asset accounting.

What BMW is using here is called an agentic AI,

even multiple of those agents between BMW and its suppliers.

First, a task agent asks the supplier to perform an inventory task.

The supplier takes a photo of the QR code

and this data runs through a validation agent that checks if everything looks correct.

If not, a human gets alerted.

If everything looks good, a documentation agent takes care of the paperwork.

The agents are based on large language models like ChatGPT,

Gemini and Claude, but they work a bit differently.

We say, hey chat GPT,

please give me three recipes that I can cook this week for myself.

And then it will do that from whatever information it has in its system.

All right, let's get cooking.

Quick, colorful and full of flavor.

An agentic AI, however, would ask for my taste preferences,

my caloric needs, habits, scrape supermarkets for cheaper offers,

check how potential items would be ranked against WHO health recommendations,

verify its own decisions and create a shopping list tailored to the layout of the supermarket that I'll be going to.

Simplifying it, an agentic AI does more with less human input and makes decisions on its own within set limits.

You can automate more and need less workers.

In narrow scenarios, it can even operate without ongoing human input.

In BMW's AI automation process,

QR codes have an important role.

This QR code is in the middle of everything.

The supplier gets the inventory order,

then he goes to the tool and scan the QR code.

Ah, so you literally just take the phone and then you take a photo and then you have all the information.

And we get all the geolocation and metadata through the photo and have the inventory number with the QR code.

That's it.

By using AI, BMW wants employees to be able to focus on cases outside the norm,

but not spend time with the ones where everything is clear.

But such an integration into an existing workflow takes time.

For BMW, it took one and a half years to get kick-started on Agentic AI.

And this integration challenge is often underestimated in the hype of things.

This technology relies, for it to be potent,

it relies on redesigning the way those employees do their work.

And that ends up being a business problem, not a technology problem.

This is David Crawford.

He helps enterprises integrate AI into their workflows.

Unlike some more recent technology disruptions,

you can't just deploy the technology and expect to get efficiency.

You actually have to do the hard work of mapping out your business process,

changing how you want to do it differently,

etc. If successful, Crawford says,

companies can see a rise in revenue of up to 10 to 25% before taxes.

But during my research...

I came across a lot of companies that promised the use of AI or agentic AI.

But when I asked them about showcasing it,

many were like, hmm, we're still testing,

product isn't ready, all sorts of excuses.

Which makes me wonder again is all of this AI stuff a bit of a*****.

And it's not just me,

there are also some well-known over promises of AI.

Do Not Pay claimed that it could automate anything a lawyer does.

The founder even offered $1 million to any lawyer

that would argue in front of the Supreme Court with the help of Do Not Pay.

Eventually, they got a fine from the Federal Trade Commission for False Promises.

In other lawyer replacing AIs,

Stanford University scientists who double-checked these AIs found severe hallucinations in their answers.

Sometimes the hallucination was so bad that it reversed the meaning of the laws.

And even benchmarks don't seem to match real-world performance.

Cognition's agent Devon, a software engineering AI,

had great benchmarking scores, but in real world they found

that Devon was only successful at three of the 20 tasks it was built for.

This GPT-5 for example and even Gemini 3 can do well on

or actually do better than humans on international math olympiads

but then they can fail on extremely silly questions that has never been in any benchmark.

This is Rao Kambamputi.

100

He works on fundamental problems in planning and and decision-making by AI systems.

101

There is such a thing as studying to the test.

102

And when you study to the test,

103

will you generalize beyond that?

104

If you actually happen to see specific exams,

105

the questions that will come.

106

The general, the entire premise of education has been that if you somehow do well in the test,

107

then the capabilities that you got in doing well on the test will help you generalize and handle other situations.

108

This notion of generalization is something that still remains like an Achilles heel for LLMs.

109

So it has become difficult to make a connection of how

110

good an AI performs in benchmarking to what it actually can do in a reliable way in the real world.

111

One of the biggest concerns around AI is always hallucinations.

112

How do you deal with them in these kind of automated workflows if anything goes wrong?

113

So the important part to control that is to have really strong guidelines

114

and our agents are fixed to one process step in the workflows.

115

So they have clear rules so that we can ensure no hallucination is happening

116

and we have always these testing possibilities in every single process step.

117

Wow, AI and no hallucinations and 100% reliable.

118

Sounds like a dream but I'm going to check with Rao Kambambutti later.

119

But it's certainly a fit for David Kirsch and Brent Goldfarb's analysis of 58 historical technological innovations like electricity,

120

aviation, railroad, dot-com boom.

121

They checked why some of them created bubbles and others didn't.

122

And they ended up with four reoccurring factors.

123

a narrative to buy into like we're we might have ai

124

that is smarter than any human by the end of this year um

125

and i would say no later than next year um

126

and then probably by 2030

127

or 2031 call it five years from now uh ai will

128

be smarter than uh all of humanity collectively i have engineers

129

within anthropo who say i don't write any code anymore i just i just let the model write the code

130

I edit it, I do the things around it.

131

I think, I don't know,

132

we might be 6 to 12 months away.

133

Second, uncertainty around how to use the new thing.

134

And if 95% of companies don't see a return on investment on AI,

135

that's a lot of uncertainty.

136

Third, the presence of pure-play companies,

137

the ones that do nothing but AI.

138

And fourth, novice investors.

139

So what we did is we took this set of technologies and said,

140

well, you know, and for each one,

141

we did a kind of deep dive into the history of that technology.

142

Where did it come from?

143

Who were the promoters?

144

What were the opportunities associated with it?

145

How did it change economic activity?

146

and these four pillars emerged as associated with increasing the likelihood of a bubble.

147

It's not a deterministic model.

148

We can't say, oh, there are narratives,

149

technological uncertainty, pure plays, novice investors,

150

therefore there will be a bubble.

151

And the first pointer for a bubble is narrative.

152

You and I are living through this once-in-human-history transition where humans go from

153

being the smartest thing on planet Earth to not the smartest thing on planet Earth.

154

These narratives or stories ultra-simplify complex processes,

155

and part of the reason why everyone is like,

156

yeah, that might be true,

157

is because we've been surrounded by these narratives for quite some time.

158

For example, by Hollywood movies like Star Trek,

159

Her, Space Odyssey, Axe Machina,

160

The Matrix, where technology can do almost anything.

161

This technology has extremely high narratability.

162

It can support so many narratives that,

163

let's say I don't buy the one about replacing the need for human labor,

164

but there's some other narrative that AI does kind of plug into that connects to my preferences.

165

And then all of a sudden I'm like,

166

okay, I'll write a check for that.

167

So check for the narrative factor on the AI bubble.

168

BMW also uses Agentic AI for a different task.

169

Automated offers for companies that want to buy,

170

lease a lot of cars.

171

So there are all sorts of questions about car models,

172

specifications, engine sizes, lease length,

173

rates that usually would all need to be read and categorized and answered by a human employee.

174

But at BMW's subsidiary Alphabet,

175

all these emails, inquiries get fed into an AI

176

that lays out a rough answer and creates quotes that a human double checks.

177

So what was a difficult thing to work with

178

when you need an AI to access all these kind of different data points?

179

Well, the challenge was to make the system work for the most cases,

180

most of the times.

181

So it's easy enough to make the AI make a perfect job just for one configuration.

182

But trying to do this for all of them,

183

adjusting the prompts and the settings and the configuration of the systems,

184

and then you have run it against all different documents again to see how it performs against them.

185

In AI lingo, that is called the reliability of a model.

186

The holy grail here are the five nines,

187

meaning a reliability of 99.999%.

188

Even top models today have failure rates of at least 3 but up to 10%.

189

So a lot of hallucinations when it comes to summarizing short documents.

190

And many experts even argue that these hallucinations are here to stay.

191

You might think 90% doesn't sound too bad,

192

but let's transfer that to a real world example.

193

Flights.

194

What if 10% of flights in the US would crash?

195

That would have been about 1.7 million crashes in 2025.

196

Absolutely horrific.

197

And so there is nothing that anybody can see that can theoretically remove hallucinations from such a model.

198

You can try to improve it with training,

199

but even actually there was just a recent paper,

200

I believe in nature machine intelligence,

201

talking about the fact that,

202

you know, LLMs can't reliably distinguish between fact and belief.

203

And the belief basically is,

204

if they hear somebody saying something, that becomes a belief.

205

And so there is always this question of,

206

you can try to make the data more clean,

207

or you can try to do additional sorts of training,

208

but if you just let the machine,

209

let the model come up with a long completion of a prompt,

210

there is very little reason to believe that it will have 100% accuracy.

211

But this makes AI harder to utilize for companies because there's always the looming danger of something going wrong,

212

also known as AI slop.

213

That's when an AI does a bad job and humans need to spend more time fixing their mistakes

214

than it would have taken them to do the task originally.

215

David Kirsch and Brent Goldfarb described this factor as uncertainty.

216

How will a new technology translate into real business?

217

Something that also happened when radio came along.

218

Why would we want to broadcast from one to many?

219

And the thought was, well,

220

we do that in church.

221

So therefore, it'll be used for religious services.

222

But then the question is how to pay for it.

223

Because how do you know if somebody is tuning in?

224

And this became a huge problem for early radio.

225

So big that there even was a public contest for ideas on how to make money with radio broadcasts.

226

And AI could be running into similar issues money-making wise because to scale AI,

227

you need loads of money.

228

Loads.

229

To build the necessary infrastructure, data centers.

230

Until 2030, McKinsey says, data centers alone will need $5.2 trillion in investments.

231

That's 15 times the amount of the Apollo space project costs.

232

17 space missions.

233

Just for comparison, in 2025,

234

AI tech is expected to generate around 60 billion dollars in revenue by one estimate.

235

I'm just getting dizzy from these numbers.

236

We need to find uses for this technology that transcend employee productivity

237

and actually do some of those really novel things accelerate drug discovery,

238

accelerate, you know, health and wellness, accelerate healthcare.

239

And the AI industry needs to find new revenue streams ASAP because they might be living on borrowed time as...

240

Even companies like Google, Microsoft,

241

Amazon and Meta, who've got a lot of cash sitting around that they can spend on data centers,

242

even they are taking on debt to finance their data centers.

243

In one of the largest private debt deals on record,

244

Meta received 72 billion US dollars to finance its gigantic data center Hyperion,

245

which is going to be built in Louisiana.

246

Also, Oracle took on 18 billion in fresh debt for its AI infrastructure.

247

Until now, it was one of the main things that separated the AI boom from what it's often compared to,

248

the dotcom boom and bust.

249

Back then many companies racked up debt as they raced to lay fiber optic cables into the ground

250

that years later would become today's high-speed internet.

251

But we all know what happened before.

252

The dot-com crash.

253

And not everybody is turning a blind eye to this.

254

In October 2025 the UK's central bank,

255

Bank of England,

256

put out a warning about the increasingly debt-financed AI infrastructure

257

and AI stocks valued at close levels seen at the peak of the dot-com bubble.

258

And there's a really interesting graph in this report, this one.

259

It shows the exact moment what happened to AI stocks

260

when the news broke that the Chinese model DeepSeq can achieve similar results like ChetGPT,

261

Claw, Gemini, with a fraction of the computing power.

262

And as you can see here,

263

everything took a bit of a nosedive.

264

The stocks recovered after some time,

265

but a breakthrough like that could render these gigantic data centers and the expected revenue from them somewhat useless.

266

So that's definitely a check for uncertainty.

267

The last bubble sign from Brent and David has something to do with all of us.

268

Because people, while not really having a clue what's going on,

269

still throw money into AI stocks.

270

In 2024, 30 billion US dollars was poured into Nvidia by retail investors and made it the most popular retail stock.

271

But Brent and David even go beyond this definition of novice investors.

272

We're all novices.

273

That's the take-home lesson.

274

There was a saying on Wall Street that you're a novice until you go through a bear market.

275

Just to recognize that that can really happen.

276

Nobody's done AI before.

277

I don't know exactly how useful it's going to be in each application.

278

I don't know how expensive it's going to be over time.

279

I don't know how, we don't know actually what business model's gonna be the most effective way for AI.

280

So where does this all leave us on the bubble thing?

281

Undeniably, AI has an impact.

282

David Crawford said some results from Bain & Company clients range from 10 to 25% uplift in their earnings before taxes.

283

And BMW says...

284

We see this percentage also within the inventory workflow and

285

when we think about how we can scale now over all of our processes,

286

I think Gen.ai will be able to bring that potential.

287

But you don't get these results by throwing ChatGPT or Co-Pilot into a company.

288

Also undeniable, there is some trash talk going around,

289

some lofty company valuations and we checked all the bubble boxes from David Kirsch and Brent Goldfarb.

290

But is it March 2000,

291

just before the burst of the dotcom bubble?

292

We don't know.

293

And according to many experts, no one really does.

294

I'm using AI models as much as I can to see what they are capable of,

295

and the advancements in the last years have been

296

but I'd say that we're missing a breakthrough in reliability to match the crazy data center spending and devaluations.

297

Microsoft's stock just took a 12% nosedive because it became clear how much they spend on AI and data centers.

298

The whole AI house of cards seems to be pretty fragile and maybe we're just one deep-seek moment away from everything going...

TRENDING

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법

쉐도잉(Shadowing)은 원래 전문 통역사 훈련을 위해 개발된 언어 학습 기법으로, 다언어 학자인 Dr. Alexander Arguelles에 의해 대중화된 방법입니다. 핵심 원리는 간단하지만 매우 강력합니다: 원어민의 영어를 들으면서 1~2초의 짧은 지연으로 즉시 소리 내어 따라 말하는 것——마치 '그림자(shadow)'처럼 화자를 따라가는 것입니다. 문법 공부나 수동적인 청취와 달리, 쉐도잉은 뇌와 입 근육이 동시에 실시간으로 영어를 처리하고 재현하도록 훈련합니다. 연구에 따르면 이 방법은 발음 정확도, 억양, 리듬, 연음, 청취력, 말하기 유창성을 크게 향상시킵니다. IELTS 스피킹 준비와 자연스러운 영어 소통을 원하는 분들에게 특히 효과적입니다.

☕ 커피 한 잔 사주기

ShadowingEnglish는 여러분의 지원 덕분에 100% 무료로 운영됩니다. 서버 및 AI 유지 비용이 많이 듭니다. 여러분의 커피 한 잔이 큰 힘이 됩니다! ��

PayPal로 기부하기

ShadowingEnglish.com – 영어 쉐도잉 연습

쉐도잉(Shadowing)으로 원어민처럼 영어를 말하세요. YouTube 영상을 한 문장씩 듣고 따라 말하며, 진짜 발음과 유창성을 키우세요. IELTS 학습자들이 선택한 방법.

쉐도잉 연습: Is AI the next dot-com crash? | Business Beyond - YouTube로 영어 말하기 배우기

인기 동영상

왜 이 비디오로 영어 말하기 연습을 해야 할까요?

문맥 내 문법 및 표현

일반적인 발음 함정

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법

쉐도잉 연습: Is AI the next dot-com crash? | Business Beyond - YouTube로 영어 말하기 배우기

앱 다운로드

왜 이 비디오로 영어 말하기 연습을 해야 할까요?

문맥 내 문법 및 표현

일반적인 발음 함정

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법